Emotional, sensory and intellectual visitor experience under augmented reality mediation at the Templo del Sol, Eliécer Silva Celis Archaeological Museum (Sogamoso, Colombia): Likert ratings and open-response corpus for lexical analysis

This dataset contains the complete response set of a quasi-experimental study measuring the emotional, sensory, and intellectual dimensions of visitor experience at the Templo del Sol gallery of the Eliécer Silva Celis Archaeological Museum (MAESC), an in situ Muisca archaeological site in Sogamoso, Boyacá, Colombia. The site operates under an archaeological management plan approved by the Instituto Colombiano de Antropología e Historia (ICANH) that prohibits physical entry to the Temple, so the visitor encounter is restricted to distant visual observation.

Sixty first-time adult visitors were assigned to two conditions of thirty: an experimental group that toured the gallery with a tablet-delivered augmented reality (AR) application providing an overlaid informational layer (explanatory audio, video, music and object-level descriptions), and a control group that toured without it. Both groups completed the same 31-item post-visit instrument immediately after the tour. Data were collected between 6 and 29 May 2021.

The instrument pairs each closed Likert item with an open-ended prompt asking the respondent to justify the rating. The dataset therefore supports two distinct analytical routes: a non-parametric between-group comparison of the ordinal ratings (Mann–Whitney U), and a lexical analysis of the open responses treated as condition-partitioned corpora. The latter was performed with KH Coder using correspondence analysis of the word × external-variable table, with chi-square-based selection of the top 60 terms and a minimum term frequency of 2.

What this is
Complete response set of a quasi-experimental study measuring emotional, sensory, and intellectual visitor experience at the Templo del Sol gallery of an in situ Muisca archaeological site. Sixty first-time adult visitors toured the gallery; thirty did so with a tablet-delivered augmented reality (AR) application overlaying explanatory audio, video, and object-level content, and thirty without it. Both groups completed the same 31-item post-visit instrument, which pairs each Likert item with an open-ended prompt asking the respondent to justify the rating.

Data were collected between 6 and 29 May 2021, under capacity and mobility restrictions imposed by the COVID-19 emergency and a period of national protest, which made non-random assignment and individually scheduled visits the only feasible protocol.

The dataset supports two analytical routes: non-parametric between-group comparison of the ordinal ratings (Mann–Whitney U), and correspondence analysis of the open responses treated as condition-partitioned lexical corpora, performed in KH Coder.

Files
File    Rows    Cols    Contents
01_analysis_matrix.csv    60    41    Analysis-ready matrix, one row per participant. The file from which the reported results are computed.
02_corpus_khcoder.csv    30    28    The open responses restructured as condition-paired columns for direct import into KH Coder or IRaMuTeQ.
03_raw_form_export.csv    64    36    Google Forms export, retained for provenance. Includes the four responses excluded from analysis.
codebook.md    —    —    Variable names, value sets, scale anchors, Spanish–English glossary.
CHANGES_v2.md    —    —    What changed from version 1.0 and why.
All files are UTF-8, comma-separated, with quoted text fields that may contain line breaks. Missing values are empty cells, not sentinel codes.

Variable naming
01_analysis_matrix.csv uses a fixed three-column pattern per experiential item:

<ITEM>khcoder — the rating as a single underscored token (Muy_Feliz, Me_gusta_mucho, Aprendí_mucho), the form required for use as an external variable in KH Coder.
<ITEM> — the same rating as a human-readable label.
D_<ITEM> — the participant's free-text justification of that rating.
Prefixes: EE_ emotional experience (FELIZ, CURIOSO, ABURRIDO, CANSADO, MOLESTO); ES_ sensory experience (ESCUCHADO, OLOR, VISTO, TOCADO); EI_ intellectual experience (APRENDIZAJE). Five further open items record recall of specific gallery content with no paired scale: EI_OFRENDA, EI_PIEZA_ORIGINAL, EI_COLUMNA, EI_ASTRONOMIA, EI_FUNDADOR.

In 02_corpus_khcoder.csv the *VARIABLE prefix marks an external variable, following the IRaMuTeQ convention that KH Coder also accepts. Within each column pair, the u suffix (*APP_FELIZu) holds AR-group responses and the unsuffixed column (*APP_FELIZ) holds control-group responses.

De-identification
No direct identifiers were ever collected. The following generalisations and removals were applied to this release; see CHANGES_v2.md for the rationale.

Participant identifiers. Replaced with P01–P60. The source file numbered participants 1–30 within each group, so its identifiers were not unique across the dataset.
Age. Exact age removed; EDAD_RANGO reports bands (18-29, 30-44, 45-59, 60+).
Residence. DEPARTAMENTO and CIUDAD replaced by RESIDENCIA (Sogamoso / Otro municipio de Boyacá / Fuera de Boyacá). This also resolves the inconsistent spellings present in the source.
Education. Collapsed to four levels in NIVEL_EDUCATIVO_AGR, merging the single-respondent categories.
Timestamps. The submission timestamps in the raw export have been deleted. Resolved to the millisecond, they could be joined with museum entry records to identify individuals.
Consent item. The administrative consent checkbox has been dropped; it is not analytical.
Third-party names. Two museum staff members were named by participants in three responses. Those spans are replaced with [ANTROPÓLOGO], [GEÓLOGO], and [GUÍA] (nine substitutions in total). The archaeologist Eliécer Silva Celis, a deceased public figure and the subject of one survey item, is retained throughout.
All 730 free-text segments were read before release. No participant identifies themselves, a companion or their institution anywhere in the corpus.

Linking the raw export to the analysis matrix
The source workbook carries no shared key between the Google Forms export and the analysis matrix, so the correspondence in 03_raw_form_export.csv was reconstructed by matching open-text responses. The INCLUIDO_EN_ANALISIS column reports the outcome honestly:

Value    n    Meaning
Sí    57    Linked to a participant in 01_analysis_matrix.csv; see PARTICIPANTE.
No: respuesta incompleta    4    Twelve or fewer of 36 fields completed. Three are AR-group responses carrying demographics only; one is a blank row flagged in the source as "Valores incompletos o para revisar". These are the cases that reduce 64 submissions to 60 analysed participants.
No determinado    3    Almost certainly among the 60 analysed, but the open-text fields differ slightly between the two sheets, so the link could not be established with certainty.
Users reproducing the reported results should work from 01_analysis_matrix.csv, which is authoritative. The raw export is provenance, not a second copy of the analysis set.

Known data-quality features
These are documented rather than silently corrected, so that reproductions match the published article.

Mixed Likert coding in the raw export. Items a, c, e, g and i mix numerals (2, 3, 4) with verbal labels (Muy Feliz, Nada Aburrido) in the same column, indicating that two versions of the form were in circulation during collection. 01_analysis_matrix.csv resolves this to a single verbal scale.
Reverse-worded items are not reverse-scored. For EE_ABURRIDO, EE_CANSADO and EE_MOLESTO a higher value denotes more of the named emotion, not a better experience. Recode before computing any composite.
One age correction. See CHANGES_v2.md.
Item-level missingness. Ratings are complete for all emotional items and for ES_ESCUCHADO (60/60). Elsewhere: ES_OLOR 58/60, ES_VISTO 59/60, ES_TOCADO 56/60, EI_APRENDIZAJE 57/60. Open-response completeness ranges from 49/60 to 57/60. Where the associated article reports a single N per variable, that value is the smaller of the two group counts, which is the effective N for the Mann–Whitney comparison; the per-group counts in this file reproduce those values exactly.
Screening item. The raw export's item 1, "¿Ha ingresado alguna vez al museo Arqueológico Eliecer Silva Celis?", is answered "Sí" by every valid respondent, while the study describes its sample as first-time visitors. The item appears to have been understood as "are you entering the museum today", and the first-visit criterion was applied at recruitment. Users should not read this column as a record of prior visits.
Reuse
The dataset is self-contained and requires no proprietary software. 01_analysis_matrix.csv supports replication of the ordinal comparisons in any statistical environment; 02_corpus_khcoder.csv can be imported directly into KH Coder or IRaMuTeQ to regenerate the correspondence biplots, using chi-square selection of the top 60 terms and a minimum term frequency of 2.

All text is Spanish. The associated article reports English translations of the extracted terms and states so explicitly. Reproductions should work from the Spanish originals, since lemmatisation and stop-word behaviour are language-dependent.

Ethics
Collected under written informed consent, following the Belmont Report (1978), Colombian Resolution 8430 of 1993 and Law 1090 of 2006. All participants were adults who consented for themselves. Participation was voluntary, withdrawable at any point without consequence, and uncompensated beyond waived museum admission. The study was classified as minimal risk.